The dimensions of an MR head are becoming smaller and the tolerances are not reducing as quickly as the dimensions. In the design of a typical recent device, the tolerances on the MR stripe height represent a .+-.33% change, or with respect to the ratio of the highest to the lowest, represents a 2:1 ratio. Further, the tolerance of the width of the MR stripe (length in the direction of current flow) is .+-.20%. The tolerance of the thickness is .+-.10%. If these are considered independent variations, the total variation in the resistance of the element is about .+-.40%, or a high to low ratio of 2.33:1.
A problem resulting from the large dimensional variability is that with the normal biasing method, a large difference in power dissipation occurs with different heads within a device. In addition, current density varies significantly as the cross-section for the current (stripe height by thickness) also varies by a large amount. The basic failure mechanism is considered to be electromigration. Product life is inversely related to the cube of the current density, and exponentially to temperature (hot being bad). Since the normal biasing method uses a DC current that is fixed for all heads, the low stripe heights and thin layer thickness result in higher resistance and higher current density. The resulting power dissipation causes significantly more temperature rise than associated with a high and thick stripe. Thus temperature and current density compound and cause a much shorter life expectancy for the low stripe height and thin MR elements, compared to the higher and thicker elements.
Another consideration is that all the factors that make the resistance higher also make the the signal level higher. Thus the best signal to noise ratio occurs with the highest resistance heads. Thus, low stripe heights, thin layers and wider stripes cause good signal to noise ratios, while high, thick and narrow stripes produce poorer signal to noise ratios. A fixed bias current must then be a compromise between good signal and short life.
Another problem is that a good electronic signal to noise ratio depends upon preamplifier design. Preamplifier design has significant limits due to the reduction of voltage available and reduced power goals. Present designs typically have a single +5 volt supply (.+-.5%). The variation in the resistance of the head also provides limits on the amount of bias current that can be used due to the voltage drop across the head and leads. This is limited by the multiple resistances and active elements that must split the available voltage to the amplifier. If too much current is run through a head of high resistance, the preamplifier stage will saturate and distort the signal, causing a degradation in performance.
A still further problem is the gradual resistance increase phenomenon (GRIP) which is related to the leads within the MR head. Recent data indicates this to be greater than previously expected with the addition of several ohms to the resistance over the life of the device. Thus, a design that is too close to the limit at manufacture may induce a saturation in the amplifier as the resistance increases late in product life, causing significant performance loss.